Executive brief
AIIR is a tool used to verify and enforce security policies on software receipts and artifacts. Several verification gates could incorrectly report that unsigned or policy-violating content was verified, potentially allowing malicious or non-compliant artifacts to bypass security controls. An attacker or non-technical user could unknowingly accept untrusted content that should have been rejected, undermining the integrity of the verification process.
Technical details
The vulnerability consists of multiple fail-open logic errors in AIIR's verification and policy enforcement paths (CWE-347: Improper Verification of Cryptographic Signature, CWE-636: Not Failing Securely). Specific affected components include: (1) the `require_signing` policy gate accepting forgeable or empty fields without valid signatures; (2) a CI verification path reporting success regardless of underlying verification result; (3) a release-verification gate advertising unenforced policy limits; and (4) signature-verification paths being silently skipped for certain input categories. All four paths return success without performing actual validation. Exploitation requires an attacker or user to rely on the affected gate for security decisions; the vulnerability does not forge valid signatures or compromise correctly-signed receipts. The attack vector is network-accessible and requires no authentication or user interaction. All paths are fixed in version 1.7.0 with regression tests.
Affected products
- Invariant Systems AI AIIR < 1.7.0
Timeline
- 2026-08-28: disclosed: Vulnerability published as GHSA-73p9-6hrp-8qhr
- 2026-08-28: patched: Fixed in version 1.7.0